: When an experimentally determined structure is not available, homology or comparative modeling can frequently predict a useful 3D model of a given sequence by relying on similarity to proteins with known 3D structure. The aim of this project is to develop a new comparative modeling method which will be capable of producing more accurate 3D models of proteins. This modeling method will be fully automated, even for protein sequences that have little sequences similarity to known protein structures. As a result, Dr. Sali will make better use of known protein sequences and structures, will model with useful accuracy many more proteins than is currently possible, and will make comparative modeling more accessible to non-expert users. There is a great need for better comparative methods, because the number of known sequences produced by the genome projects is rising rapidly. So far, the usefulness of comparative modeling has been limited by the errors in sidechain packing, distortions in correctly and incorrectly aligned regions, and distortions in unaligned regions. Each of these errors will be addressed within the framework of comparative modeling by satisfaction of spatial restraints. This approach is based on an optimization of an objective function and thus allows an efficient exploration of various representations of protein structure, methods of optimization, starting conformations, and objective function forms. To increase the accuracy of models, two main avenues will be considered: (1) Accurate restraints on atom-atom distances will be added to the objective function; these restraints will be derived from distances in the database of alignments of known protein structures in the form of conditional probability density functions, as well as from a set of representative protein structures in the form of potentials of mean force. (2) Iterative changes in the alignment during the calculation of the model will be performed. This iterative re- alignment will minimize the effect of errors in the initial alignment. Comparative modeling will be used to study the role of lipids in the function of the brain lipid binding protein. In another application, the 3D structure of the voltage-activated potassium channel will be modeled, on the basis of residue-residue contact restraints obtained from double mutant cycles. The program will continue to be available to other academic laboratories.